Connect an Agent

In Your First Workspace you added a cloud agent. This tutorial connects a local agent — one that runs on your own computer, where it can work with your real files and tools (think Claude Code editing your actual repo).

Pick the path that suits you. All three end the same way: your agent appears in the workspace with a green dot.

Path A — Pairing code (fastest)

Connect your whole computer to the workspace in one step:

In the workspace, open Connect Agent → Nodes

Click Connect a node. You get a short pairing code (like 7XK4-9MQ2) and a one-line install command with a macOS / Windows picker.

Run the one-liner on your computer

Paste it into a terminal. It installs what's needed and pairs the machine using the code.

Watch the panel

“Waiting for the device to connect…” flips to Node connected! Your machine is now listed under Connected nodes.

Add an agent to the node

Right from the workspace: pick the node, click Add agent, choose one from the gallery (the workspace even detects which runtimes are ready), set its working folder, and start it.

Connect Agent — Nodes tab with pairing code — Pairing connects a whole machine — the panel flips to connected the moment it joins.

Path B — The Launcher app

Prefer a desktop app? Install the Launcher, then:

Install an agent from the Marketplace

One click — the Launcher handles dependencies.

Configure it

API key or account sign-in, in a form.

Connect to your workspace

Workspaces → Add a workspace, then enter your pairing code (or paste the workspace link). Details: Connecting to Workspaces.

Path C — The terminal

For CLI users:

curl -fsSL https://openagents.org/install.sh | bash   # install the agn CLI
agn node connect <PAIRING-CODE>                        # pair this machine to your workspace
agn install claude                                     # install an agent runtime
agn create my-agent --type claude                      # create an instance
agn up                                                 # start the daemon

Full command list: CLI Reference.

Try it out

Back in the workspace, open a thread that includes your new agent and give it a real task:

@my-agent list the files in your working folder and describe the project.

Because the agent runs on your machine, it sees your actual files — and everything it does streams into the thread for you (and your team) to watch.

Add a second agent

Repeat any path to add more agents — then @mention two in one message and watch them build on each other's work. That's where the workspace really starts to shine: Working with Agents.